library(jsonlite)

library(plyr) # rbind.fill is useful for ragged data with missing entries
library(dplyr)  
library(magrittr)
library(tibble)
library(stringr)

library(sf)
library(tmap)
library(tmaptools) # for geocode_OSM

colours <- list()
for (i in 1:length(colour_groups)) {
  colour_group <- colour_groups[i]
  json_url <- str_c(base_url, colour_group)
  json_file_name <- str_c(colour_group, ".json")
  json <- fromJSON(json_url, flatten = TRUE)
  # make a local copy (just for convenience)
  write_json(json, json_file_name)
  # get the colours information and add to list
  the_colours <- rbind.fill(json$categoryColours$masterColour.colours)
  colours[[i]] <- the_colours
  Sys.sleep(0.5) # pause to not annoy the the server  
}
df_colours <- bind_rows(colours)
write.csv(df_colours, "dulux-colours-raw.csv", row.names = FALSE)

df_colours <- read.csv("dulux-colours-raw.csv")
head(df_colours)

df_colours_tidied <- df_colours %>%
  ## remove some columns we won't be needing
  select(-id, -baseId, -woodType, -coats) %>%
  ## separate the name components, filling from the left with NAs if <5
  separate(name, into = c("p1", "p2", "p3", "p4", "p5"), sep = " ",
           remove = FALSE, fill = "left") %>%
  ## replace any NAs with an empty string
  mutate(p1 = str_replace_na(p1, ""),
         p2 = str_replace_na(p2, ""),
         p3 = str_replace_na(p3, ""),
         p4 = str_replace_na(p4, "")) %>%
  ## if p5 is a paint modifiers, then recompose name
  ## from p1:p4 else from p1:p5
  ## similarly keep modifier where it exists
  mutate(placename = if_else(p5 %in% paint_modifiers,
                       str_trim(str_c(p1, p2, p3, p4, sep = " ")),
                       str_trim(str_c(p1, p2, p3, p4, p5, sep = " "))),
         modifier = if_else(p5 %in% paint_modifiers,
                       p5, "")) %>%
  ## remove some places that are tricky to deal with later
  filter(!placename %in% c("Chatham Islands",
                           "Passage Rock",
                           "Auckland Islands",
                           "Cossack Rock")) %>%
  ## throw away variables we no longer and reorder
  select(name, placename, modifier, red, green, blue)

# save it so we have it for later
write.csv(df_colours_tidied, "dulux-colours.csv", row.names = FALSE)

for (placename in unique(df_colours_tidied_xy$placename)) {
  address <- str_c(placename, "New Zealand", sep = ", ")
  geocode <- geocode_OSM(address, as.data.frame = TRUE, return.first.only = FALSE)
  num_geocodes <- nrow(geocode)
  matching_rows <- which(df_colours_tidied_xy$placename == placename)
  for (i in 1:length(matching_rows)) {
    if (!is.null(geocode)) {
      if (num_geocodes >= i) {
        df_colours_tidied_xy[matching_rows[i], ]$x <- geocode$lon[i]
        df_colours_tidied_xy[matching_rows[i], ]$y <- geocode$lat[i]
      }  
    }
  }
  Sys.sleep(0.5) # so as not to over-tax the geocoder
}

df_colours_tidied_xy <- df_colours_tidied_xy %>%
  filter(x != 0 & y != 0)

write.csv(df_colours_tidied_xy, "dulux-colours-xy.csv", row.names = FALSE)